Interleaved EPI diffusion imaging using SPIRiT‐based reconstruction with virtual coil compression
For high‐resolution DWI, one choice is to use multishot acquisition techniques 12, which can increase the PE bandwidth to reduce the geometric distortion. However, multishot DWI suffers from phase variations among different shots due to minuscule motions during the application of diffusion encoding gradients. If not corrected, these phase variations can cause severe ghost artifacts in final DW images. Developing robust multishot sequences with accurate correction methods has been an ongoing effort. Currently, the methods generally are divided into three categories: extra‐navigated methods 13, self‐navigated methods 17, and navigator‐free methods 20.
Recently, a k‐space–based reconstruction method was proposed for 2D navigated interleaved EPI (iEPI) DWI 25. Instead of removing the phase variations from each shot in the image domain 14, channel‐by‐channel autocalibrating reconstruction in the k‐space domain using generalized autocalibrating partially parallel acquisitions (GRAPPA) 11 was applied to calculate the diffusion image from each shot and each coil. This new method is less sensitive to mismatches between navigator and image echoes than image domain‐based methods 28; thus, strict requirements about the navigator acquisition with minimal mismatches can be alleviated 29.
As a new approach to autocalibrating, kernel‐based parallel imaging reconstruction, iterative self‐consistent parallel imaging reconstruction (SPIRiT) 30 improves the performance of traditional GRAPPA‐like methods in that the formulation of iteratively enforcing self‐consistency is better conditioned. These features suggest that SPIRiT can be a better candidate for parallel imaging reconstruction, especially when the acceleration factor is pushed to a high limit. In this work, we extend SPIRiT to integrate coil and shot dimensions to further improve the robustness of multishot DWI reconstruction, and apply it to 2D navigated iEPI DWI. In addition, when the number of coil arrays and the number of shots are large, the reconstruction becomes very computationally intensive. Coil compression can reduce redundant coil information and accelerate the computation. Thus, we integrate coil compression 31 with SPIRiT and propose a novel virtual coil compression method, shot‐coil compression, for further computation acceleration. In brief, the proposed method combines the SPIRiT‐based reconstruction and shot‐coil compression to achieve robust reconstruction for multishot DWI.